This article presents the results of a qualitative study on older adults’ first impressions of the ChatGPT 4.0 voice interface in the Polish language. We organized three exploratory workshops with older adult volunteers (N = 13), during which we inquired about their expectations and observed their first actual use of a fluent conversational voice interface in Polish. Based on thematic analysis of acquired data we classified recurring themes, impressions, aspirations and opportunities for improvement. In this exploratory work we formulate a list of nine preliminary insights to consider when designing voice interaction systems for older adults using large language models. Our study serves as a unique snapshot of a non-English-speaking population’s first interaction with a fluent, non-rule-based voice assistant in their native language, and forms a valuable baseline for future work in Human–AI interaction.